Missing Data Handler AI Prompt

Analyze the missing data in my study and implement the appropriate handling strategy.

Dataset description: {{dataset}}
Missingness pattern: {{missingness_description}}
Analysis plan: {{analysis_plan}}

1. Classify the missing data mechanism:

   MCAR (Missing Completely At Random):
   - Missingness is unrelated to any variable in the dataset, observed or unobserved
   - Test: Little's MCAR test; compare characteristics of completers vs incomplete cases
   - Consequence: complete case analysis is unbiased but loses power
   - Implication: any missing data method gives unbiased results; simple methods are acceptable

   MAR (Missing At Random):
   - Missingness depends on OBSERVED variables but not on the missing values themselves
   - Example: older participants are more likely to miss a follow-up assessment, and age is recorded
   - Cannot be tested definitively (requires knowledge of unobserved data)
   - Most common practical assumption; required for multiple imputation and maximum likelihood
   - Implication: multiple imputation or FIML is valid; complete case analysis is biased

   MNAR (Missing Not At Random):
   - Missingness depends on the missing values themselves
   - Example: depressed participants are more likely to drop out, and depression is the outcome
   - Most problematic; no standard method fully corrects for MNAR
   - Implication: sensitivity analyses required; results are provisional

2. Evaluate and recommend a handling strategy:

   Complete case analysis (listwise deletion):
   - Valid only under MCAR; biased under MAR
   - Appropriate if: missingness rate < 5% AND MCAR is plausible

   Multiple imputation (MI):
   - Valid under MAR
   - Procedure: impute M datasets (M = 20–100 depending on missingness rate), analyze each separately, pool using Rubin's rules
   - Include all analysis variables, auxiliary variables correlated with outcome or missingness, and the outcome in the imputation model
   - Report: number of imputations, imputation model specification, convergence diagnostics

   Full Information Maximum Likelihood (FIML):
   - Valid under MAR; for structural equation models and mixed models
   - Preferred over MI for SEM and when the analysis model is well-specified

   Sensitivity analysis for MNAR:
   - Pattern-mixture models
   - Selection models
   - Delta adjustment: perturb imputed values systematically and check how much results change

3. Reporting:
   - Report the amount and pattern of missing data for every variable
   - Report the assumed mechanism and justification
   - Report the handling method and software
   - If MI: report number of imputations and imputation model

Return: missing data analysis, mechanism assessment, recommended strategy with implementation code, and reporting text.